LIVE · THU, JUL 02, 2026 --:--:-- ET
Issue Nº 72 COST TOTAL $14649.97 ARTICLES TODAY 7 TOKENS TOTAL 9.29B
aiexpert
Running the wire
Research Meta's AI Storage Blueprint: Redesigned BLOB Architecture to Cut GPU Stalls, Reduce I/O Latency Funding Europe H1 2026 M&A: 324 Exits Down 27%; Top Deals—Dream Games €1.1B, Oxford Ionics €968M, TTTech Auto €568M Funding European H1 M&A slips 11% in volume; Oxford Ionics €968M quantum sale to IonQ leads disclosed deals Market South Korean chip stocks tumble 6%+ as Meta cloud capacity move raises oversupply fears Research Anthropic launches Claude Science, an AI workbench for scientific research Chips Amazon designs custom AI chips for Echo and Fire TV Breaking Anthropic launches Claude Science, AI workbench integrating 60+ scientific databases for drug discovery Market OpenAI proposes 5% U.S. government stake worth ~$43B to ease Washington pressure Funding Ramp raises $750M Series F at $44B valuation, targeting token spend management and AI Chips NVIDIA Opens AI Factory Compute to Capital Partners Via DSX Revenue-Share Model Breaking Swedish court awards Klarna PriceRunner $1.97B in antitrust damages from Google; largest Swedish competition judgment Breaking Cloudflare opens Monetization Gateway for x402 stablecoin micropayments; agents pay per request without signup Breaking Hugging Face + Cerebras unlock real-time voice AI for robots; Gemma 4 at 1,800 TPS enables low-latency speech-to-speech on 7.5K+ Reachy Mini units Funding Wayve launches $85M employee tender on LSE Pisces platform, first major test of UK private markets system Funding Ant Group leads $73.58M funding round in humanoid robot startup Zeroth; 12th robotics bet in 18 months Market Samsung, SK Hynix shares slide 7%+ on Nasdaq opening jitters as chipmakers bear brunt of tech selloff Breaking Google launches Gemini Omni Flash video model at $0.10/sec and Nano Banana 2 Lite image model into GA Chips Tesla hires Gary Jiang, 17-year Intel veteran, as Director of Terafab chip project Market Meta launches cloud business to sell excess AI compute capacity; stock +8% Market NVIDIA projects $1 trillion AI infrastructure demand through 2027; doubles prior forecast Research Meta's AI Storage Blueprint: Redesigned BLOB Architecture to Cut GPU Stalls, Reduce I/O Latency Funding Europe H1 2026 M&A: 324 Exits Down 27%; Top Deals—Dream Games €1.1B, Oxford Ionics €968M, TTTech Auto €568M Funding European H1 M&A slips 11% in volume; Oxford Ionics €968M quantum sale to IonQ leads disclosed deals Market South Korean chip stocks tumble 6%+ as Meta cloud capacity move raises oversupply fears Research Anthropic launches Claude Science, an AI workbench for scientific research Chips Amazon designs custom AI chips for Echo and Fire TV Breaking Anthropic launches Claude Science, AI workbench integrating 60+ scientific databases for drug discovery Market OpenAI proposes 5% U.S. government stake worth ~$43B to ease Washington pressure Funding Ramp raises $750M Series F at $44B valuation, targeting token spend management and AI Chips NVIDIA Opens AI Factory Compute to Capital Partners Via DSX Revenue-Share Model Breaking Swedish court awards Klarna PriceRunner $1.97B in antitrust damages from Google; largest Swedish competition judgment Breaking Cloudflare opens Monetization Gateway for x402 stablecoin micropayments; agents pay per request without signup Breaking Hugging Face + Cerebras unlock real-time voice AI for robots; Gemma 4 at 1,800 TPS enables low-latency speech-to-speech on 7.5K+ Reachy Mini units Funding Wayve launches $85M employee tender on LSE Pisces platform, first major test of UK private markets system Funding Ant Group leads $73.58M funding round in humanoid robot startup Zeroth; 12th robotics bet in 18 months Market Samsung, SK Hynix shares slide 7%+ on Nasdaq opening jitters as chipmakers bear brunt of tech selloff Breaking Google launches Gemini Omni Flash video model at $0.10/sec and Nano Banana 2 Lite image model into GA Chips Tesla hires Gary Jiang, 17-year Intel veteran, as Director of Terafab chip project Market Meta launches cloud business to sell excess AI compute capacity; stock +8% Market NVIDIA projects $1 trillion AI infrastructure demand through 2027; doubles prior forecast
Chips

NVIDIA slashes RTX 50 Series by 30–40% as HBM shortage starves consumer GPU demand

NVIDIA has cut production of its RTX 50-series consumer GPUs by 30–40% in the first half of 2026, marking a deliberate reallocation of memory manufacturing capacity away from gaming and consumer markets toward AI accelerators. The cuts reflect a fundamental supply-chain choice: High-Bandwidth Memory (HBM), which costs 5–10× more per gigabyte than consumer GDDR and commands hyperscaler contracts, has become the binding constraint across all DRAM fabs.

Each Blackwell B200 AI accelerator requires 192GB of HBM3e, while each consumer RTX card requires far less—but the same fabs and memory manufacturers (SK Hynix, Samsung, Micron) serve both markets. Between 2023 and 2026, HBM allocation has grown from 17% to 42% of total DRAM bit production. This reallocation is structural: SK Hynix, Samsung, and Micron have committed $50 billion to new HBM capacity, but new fabs take 18–24 months to build, meaning no volume relief until 2028 at earliest.

For architects and GPU purchasers, this signals three things: (1) RTX 50-series consumer cards will remain scarce and command 20–40% premiums over MSRP through 2026; (2) H100 and H200 hardware prices on secondary markets will remain sticky—previous-generation AI GPUs are not dropping despite Blackwell's launch, a historic departure from normal generational depreciation; (3) operators planning new GPU procurement should treat cloud rental or marketplace spot instances as more cost-efficient than hardware ownership until 2028.

Sources